This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Patch to fix optabs.c when SWITCHABLE_TARGETS is defined.


On 11/07/2011 09:43 AM, Tom de Vries wrote:
On 11/06/2011 04:04 PM, Andrew MacLeod wrote:
(get_atomic_op_for_code): New. Return an opcode table entry.
I'm getting these errors with a MIPS cross build:
...
/scratch/vries/b4/ref-mips-ml-11-11-07/src/gcc-mainline/gcc/optabs.c: In
function 'get_atomic_op_for_code':
/scratch/vries/b4/ref-mips-ml-11-11-07/src/gcc-mainline/gcc/optabs.c:7556:
error: initializer element is not constant
/scratch/vries/b4/ref-mips-ml-11-11-07/src/gcc-mainline/gcc/optabs.c:7556:
error: (near initialization for 'add_op.mem_fetch_before')
/scratch/vries/b4/ref-mips-ml-11-11-07/src/gcc-mainline/gcc/optabs.c:7556:
error: initializer element is not constant
...

When SWITCHABLE_TARGETS is defined, the various direct optab entries can be changed on the fly, so the direct_optab tables no longer become something that can be filled in statically at compile time.


This patch changes it so that we just fill in a table entry on demand.

bootstrapped and no new regressions on x86_64-unknown-linux-gnu in the atomic testsuite. (full testuite run still underway)

Tom does this fix the mips problem? it seems to compile here now.

Presuming the full testsuite run comes out fine, ok for mainline?

Andrew

Attachment: switchable.diff
Description: Text document


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]